linux-kernel - Re: [PATCH] perf scripting python: expose symbol offset and source information

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <3D8DDD29-A919-4E3C-8256-B7F1DC4CFFED@redhat.com>
Date:   Tue, 22 Feb 2022 16:05:26 +0100
From:   Eelco Chaudron <echaudro@...hat.com>
To:     Jiri Olsa <olsajiri@...il.com>
Cc:     peterz@...radead.org, mingo@...hat.com, acme@...nel.org,
        mark.rutland@....com, alexander.shishkin@...ux.intel.com,
        jolsa@...hat.com, namhyung@...nel.org,
        linux-perf-users@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] perf scripting python: expose symbol offset and source information



On 10 Feb 2022, at 8:52, Jiri Olsa wrote:

> On Thu, Feb 03, 2022 at 09:44:33AM -0500, Eelco Chaudron wrote:
>> This change adds the symbol offset to the data exported for each
>> call-chain entry. This can not be calculated from the script and
>> only the ip value, and no related mmap information.
>>
>> In addition, also export the source file and line information, if
>> available, to avoid an external lookup if this information is needed.
>
> could you please update Documentation/perf-script-python.txt with that?
>
> any example script under scripts/python would be great

Looks like there is no documentation at all regarding the call chains, only the auto-generate script.
So rather than trying to describe this, I decided to update the auto-generated script to include this information as an example.

The current output, if it included a callchain is:

        [ffffffff99e03f0e] syscall_trace_enter
        [ffffffff99e03f0e] syscall_trace_enter
        [ffffffff99e043a9] do_syscall_64
        [ffffffff9a8000ad] entry_SYSCALL_64_after_hwframe
        [7f5b552280db] __GI_getrusage
        [17a4d78] getrusage_thread
        [17a4d78] refresh_rusage
        [17a4d78] time_poll
        [178f683] poll_block
        [16bceb5] udpif_revalidator
        [177b71c] ovsthread_wrapper
        [7f5b57a66179] start_thread
        [7f5b55231dc2] __GI___clone

The new output will be (dependend on the information available):

        [ffffffff99e0392f] syscall_slow_exit_work+0xaf ([kernel.kallsyms])
        [ffffffff99e0392f] syscall_slow_exit_work+0xaf ([kernel.kallsyms])
        [ffffffff99e043d0] do_syscall_64+0x170 ([kernel.kallsyms])
        [ffffffff9a8000ad] entry_SYSCALL_64_after_hwframe+0x65 ([kernel.kallsyms])
        [7f5b55226a41] __GI___poll+0x51 (/usr/lib64/libc-2.28.so)
        [17a4c50] time_poll+0x190 (/usr/sbin/ovs-vswitchd) timeval.c:326
        [178f683] poll_block+0x83 (/usr/sbin/ovs-vswitchd) poll-loop.c:364
        [16bceb5] udpif_revalidator+0x185 (/usr/sbin/ovs-vswitchd) ofproto-dpif-upcall.c:1024
        [177b71c] ovsthread_wrapper+0x5c (/usr/sbin/ovs-vswitchd) ovs-thread.c:422
        [7f5b57a66179] start_thread+0xe9 (/usr/lib64/libpthread-2.28.so)
        [7f5b55231dc2] __GI___clone+0x42 (/usr/lib64/libc-2.28.so)

I will send out the v2 soon.

>>
>> Signed-off-by: Eelco Chaudron <echaudro@...hat.com>
>> ---
>>  .../util/scripting-engines/trace-event-python.c    |   42 ++++++++++++++------
>>  1 file changed, 30 insertions(+), 12 deletions(-)
>>
>> diff --git a/tools/perf/util/scripting-engines/trace-event-python.c b/tools/perf/util/scripting-engines/trace-event-python.c
>> index e752e1f4a5f0..0f392b4ff663 100644
>> --- a/tools/perf/util/scripting-engines/trace-event-python.c
>> +++ b/tools/perf/util/scripting-engines/trace-event-python.c
>> @@ -392,6 +392,18 @@ static const char *get_dsoname(struct map *map)
>>  	return dsoname;
>>  }
>>
>> +static unsigned long get_offset(struct symbol *sym, struct addr_location *al)
>> +{
>> +	unsigned long offset;
>> +
>> +	if (al->addr < sym->end)
>> +		offset = al->addr - sym->start;
>> +	else
>> +		offset = al->addr - al->map->start - sym->start;
>> +
>> +	return offset;
>> +}
>> +
>>  static PyObject *python_process_callchain(struct perf_sample *sample,
>>  					 struct evsel *evsel,
>>  					 struct addr_location *al)
>> @@ -443,6 +455,24 @@ static PyObject *python_process_callchain(struct perf_sample *sample,
>>  					_PyUnicode_FromStringAndSize(node->ms.sym->name,
>>  							node->ms.sym->namelen));
>>  			pydict_set_item_string_decref(pyelem, "sym", pysym);
>> +
>> +			if (node->ms.map) {
>> +				struct map *map = node->ms.map;
>> +				struct addr_location node_al;
>> +				unsigned long offset;
>> +
>> +				node_al.addr = map->map_ip(map, node->ip);
>> +				node_al.map  = map;
>> +				offset = get_offset(node->ms.sym, &node_al);
>> +
>> +				pydict_set_item_string_decref(
>> +					pyelem, "sym_off",
>> +					PyLong_FromUnsignedLongLong(offset));
>> +			}
>> +			if (node->srcline && strcmp(":0", node->srcline))
>> +				pydict_set_item_string_decref(
>> +					pyelem, "sym_srcline",
>> +					_PyUnicode_FromString(node->srcline));
>
> nit missing { } for multiline if code
>

Will fix in v2
>>  		}
>>
>>  		if (node->ms.map) {
>> @@ -520,18 +550,6 @@ static PyObject *python_process_brstack(struct perf_sample *sample,
>>  	return pylist;
>>  }
>>
>> -static unsigned long get_offset(struct symbol *sym, struct addr_location *al)
>> -{
>> -	unsigned long offset;
>> -
>> -	if (al->addr < sym->end)
>> -		offset = al->addr - sym->start;
>> -	else
>> -		offset = al->addr - al->map->start - sym->start;
>> -
>> -	return offset;
>> -}
>> -
>>  static int get_symoff(struct symbol *sym, struct addr_location *al,
>>  		      bool print_off, char *bf, int size)
>>  {
>>